基于术前IL-6、PGE2、TNF-α构建预测膀胱癌患者术后复发的列线图模型

Construction of visualization model for predicting postoperative recurrence of bladder cancer based on preoperative IL-6,PGE2 and TNF-αlevels

目的基于术前血清IL-6、前列腺素E2(PGE2)、TNF-α构建预测膀胱癌术后复发的列线图模型。方法回顾性收集2018年6月至2023年2月临平区第一人民医院收治的348例膀胱癌患者的临床资料,经计算机产生随机数表并以2∶1比例将其分为训练集(232例)和验证集(116例)。所有患者均接受随访,将发生复发的患者纳入复发组,未发生复发的患者纳入未复发组。比较训练集复发组、未复发组血清IL-6、PGE2、TNF-α水平及一般资料;用Logistic回归模型分析训练集膀胱癌术后复发的影响因素,并建立回归方程;用ROC曲线分析术前IL-6、PGE2、TNF-α单独及联合预测膀胱癌术后复发的效能;建立膀胱癌术后复发的风险预测列线图模型,并验证其效能。结果与未复发组比较,复发组血清IL-6、PGE2、TNF-α水平升高,肿瘤直径增大,多发性肿瘤、肿瘤分期T 2~T 4、肿瘤WHO病理学分级Ⅱ~Ⅲ级的构成比升高,术后规律膀胱灌注的构成比降低(P<0.05)。Logistic回归分析显示,术前血清IL-6、PGE2、TNF-α、肿瘤分期、肿瘤WHO病理学分级是膀胱癌术后复发的影响因素(P<0.05),并建立Logistic回归方程:Y=1.718 X 1+2.081 X 2+1.815 X 3+2.319 X 4+1.868 X 5。ROC曲线显示,术前IL-6、PGE2、TNF-α预测膀胱癌术后复发的最佳截断点分别为0.60 ng/L、57.13 pg/mL、2.10 ng/mL,三者单独及联合预测膀胱癌的ROC曲线下面积(AUC ROC)分别为0.729、0.743、0.733和0.825。基于训练集Logistic回归分析结构建立膀胱癌术后复发的风险预测列线图模型,该模型预测训练集、验证集的敏感性分别为94.12%、90.20%,特异性分别为90.06%、87.29%,AUC ROC分别为0.940、0.914;Bootstrap法内部验证结果显示,训练集、验证集的C-index分别为0.918(95%CI:0.824~0.987)、0.901(95%CI:0.835~0.957)。结论术前血清IL-6、PGE2、TNF-α水平是膀胱癌术后复发的影响因素,据此建立的风险预测列线图模型具有良好的预测效能。

Objective To construct a visualization model for predicting the postoperative recurrence of bladder cancer based on preoperative serum interleukin-6(IL-6),prostaglandin E2(PGE2),and tumor necrosis factor-alpha(TNF-α)levels.Methods The clinical data of 348 patients with bladder cancer admitted to our hospital from June 2018 to February 2023 were retrospectively collected,and they were randomly divided into the training set(n=232)and validation set(n=116)at a ratio of 2∶1 by the computer-generated random number table.All patients were followed up.The patients who had recurrence were included in the recurrence group,and those who did not have recurrence were included in the non-recurrence group.The levels of serum IL-6,PGE2,and TNF-αand general data of the recurrence group and non-recurrence group in the training set were compared.The Logistic regression model was used to analyze the influencing factors of postoperative recurrence of bladder cancer in the training set and establish a regression equation.The receiver operating characteristic(ROC)curve was used to analyze the efficacy of preoperative IL-6,PGE2,and TNF-αalone and their combination in predicting the postoperative recurrence of bladder cancer.A risk prediction nomogram model for the postoperative recurrence of bladder cancer was established and validated.Results Compared with the non-recurrence group,the levels of serum IL-6,PGE2,and TNF-α,tumor diameter,and proportions of multifocal tumors,tumor stage T 2~T 4,and WHO tumor pathological gradeⅡ~Ⅲin the recurrence group were increased,while the proportion of regular bladder perfusion after surgery was decreased(P<0.05).Logistic regression analysis showed that preoperative serum IL-6,PGE2,and TNF-αlevels,tumor stage,and WHO tumor pathological grade were the influencing factors of postoperative recurrence of bladder cancer(P<0.05).The established Logistic regression equation was:Y=1.718 X 1+2.081 X 2+1.815 X 3+2.319 X 4+1.868 X 5.The ROC curve showed that the optimal cut-off points of preoperative serum IL-6,PGE2,and TNF-αlevels for predicting the postoperative recurrence of bladder cancer were 0.60 ng/L,57.13 pg/mL,and 2.10 ng/mL,respectively.The areas under the ROC curve(AUC ROC)of serum IL-6,PGE2,and TNF-αalone and their combination were 0.729,0.743,0.733,and 0.825,respectively.Based on the structure of Logistic regression analysis in the training set,a risk nomogram model for predicting the postoperative recurrence of bladder cancer was established.The prediction sensitivity,specificity,and AUC ROC of the model in the training set and validation set were 94.12%and 90.20%,90.06%and 87.29%,and 0.940 and 0.914,respectively.The internal validation results of the Bootstrap method showed that the C-index values of the training set and validation set were 0.918(95%CI:0.824-0.987)and 0.901(95%CI:0.835-0.957),respectively.Conclusion Preoperative serum IL-6,PGE2,and TNF-αlevels are the influencing factors of postoperative recurrence of bladder cancer,and the risk nomogram model based on them has good prediction efficacy.